Abstract
Recently, the Markov state model has been applied for kinetic analysis of molecular dynamics simulations. However, discretization of the conformational space remains a primary challenge in model building, and it is not clear how the space decomposition by distinct clustering strategies exerts influence on the model output. In this work, different clustering algorithms are employed to partition the conformational space sampled in opening and closing of fatty acid binding protein 4 as well as inactivation and activation of the epidermal growth factor receptor. Various classifications are achieved, and Markov models are set up accordingly. On the basis of the models, the total net flux and transition rate are calculated between two distinct states. Our results indicate that geometric and kinetic clustering perform equally well. The construction and outcome of Markov models are heavily dependent on the data traits. Compared to other methods, a combination of Bayesian and hierarchical clustering is feasible in identification of metastable states.
Original language | English (US) |
---|---|
Pages (from-to) | 1205-1215 |
Number of pages | 11 |
Journal | Journal of Chemical Information and Modeling |
Volume | 56 |
Issue number | 6 |
DOIs | |
State | Published - Jun 27 2016 |
Bibliographical note
Funding Information:This work was supported by The Hormel Foundation and National Institutes of Health, Grants CA172457, CA166011, and R37 CA081064.
Publisher Copyright:
© 2016 American Chemical Society 2016.